CFD and machine learning in building performance simulation: towards urban microclimate integration

This paper concisely traces the evolution of Computational Fluid Dynamics (CFD) in building performance simulation, from early numerical methods to contemporary applications. Initial CFD studies often treat buildings in isolation due to hardware and software limitations, neglecting interactions with the surrounding microclimate. Today, with increasing urban density, climate change, and electrification, incorporating urban microclimate effects has become essential. Recent advances – such as Physics-Informed Neural Networks (PINNs), AI-driven methods, and IoT sensors – are improving CFD's efficiency and enabling real-time, adaptive approaches to climate-responsive design. The article summarizes some developments and offers some insights into its future within building performance simulation.